#5151st of 70 in Open models
skywork-r1v
Open-weight 38B multimodal reasoning model with inference and eval code
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 3.2k
- License
- MIT
- Last commit
- Jul 2026
- Language
- Python
3,142 on 21 Apr 20263,173 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Skywork-R1V3 is a 38B vision-language reasoning model post-trained mainly with reinforcement learning on top of InternVL3-38B. The repository holds inference scripts for Transformers and vLLM, plus evaluation code for reproducing the reported benchmarks. The README reports 76.0 on MMMU (val), and weights are hosted on Hugging Face.
Who it is for: Researchers and engineers running open multimodal reasoning models on GPUs
Strengths
- Reports 76.0 MMMU and tops MathVista, MathVerse and MMK12 in its own comparison table
- Includes eval code to reproduce the benchmark results
- Inference scripts for both Transformers and vLLM with tensor parallelism
- MIT-licensed code on an MIT-licensed InternVL3-38B base
Weaknesses
- Benchmark numbers are self-reported; some baseline scores come from the authors' own framework
- 38B size needs multiple GPUs for the full model
- Trails Claude 3.7 on EMMA, MathVision and SeePhys in the README table
- No Docker image or packaged server; setup is conda plus shell scripts
What it needs
- GPU required
- Needs conda, Python 3.10, Transformers, vLLM
- Models: Skywork-R1V3-38B, Skywork-R1V2-38B, Skywork-R1V2-38B-AWQ, Skywork-R1V-38B, Skywork-R1V-38B-AWQ
Quick answers
Is skywork-r1v open source?
Yes. Its code is on GitHub under the MIT license.
Does skywork-r1v need a GPU?
Yes, it needs a GPU.
Also in Open models
See all 70| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 79 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 76 out of 100 |